Number of standby redo log groups

Yong Huang

Eric,

How many log groups do you have on the primary? 2?

You have 3 standby log groups. If you keep switching log file groups on the primary (not by "alter system switch logfile", but by Oracle completely filling up log files with redo), do all the 3 standby log groups get used, or only 2 of them take turns to be used? Thanks.

What's your Oracle version?

Yong Huang

eagle_fan

原帖由 Yong Huang 于 2009-3-30 05:25 发表
Eric,

How many log groups do you have on the primary? 2?

You have 3 standby log groups. If you keep switching log file groups on the primary (not by "alter system switch logfile", but by Oracle completely filling up log files with redo), do all the 3 standby log groups get used, or only 2 of them take turns to be used? Thanks.

What's your Oracle version?

Yong Huang

There are four log groups on primary and each log file is 2GB. During peak time, the log file is filled in 2~3 minutes.

If the network transfer speed is not quick enough, it will write the standby log group in parallel using arc processes. All of the standby log groups can be used, not only 2.

This only happens on extremely busy database.

Oracle version is 10203.

Yong Huang

原帖由 eagle_fan 于 2009-3-30 21:07 发表

There are four log groups on primary and each log file is 2GB. During peak time, the log file is filled in 2~3 minutes.

If the network transfer speed is not quick enough, it will write the standby log group in parallel using arc processes. All of the standby log groups can be used, not only 2.

This only happens on extremely busy database.

Oracle version is 10203.

Thanks. But the key question in this discussion is whether an extra SRL group is needed. If you have 4 log groups on primary, I would like you to have 5 SRL groups (more than 5 is fine too), and see whether only 4 or all 5 on standby will be used during heavy transaction period. If you have only 3 SRL groups (based on your earlier message), then the answer to the question is still not clear.

By the way, how did you make the standby use SRLs with ARCn processes shipping redo? I know if you use LGWR to ship redo, SRLs will be used. There may be a way to use SRLs even if ARCn ship redo but I'm not sure how. You're running managed standby, aren't you?

Yong Huang

opsc

In 10g, archive process will do the archive to SRL immediately if the LNS process is busy. I can't figure out how oracle controls the LNS process number, but in our case, it's at most one LNS active. In 9i, there is some threshold like timeout setting for log_archive_descX lgwr option, so this may make a bit difference.

  So the question you pointed out is easy to answer, we can have up to 30 archive process working on the transfering primary redo,  so simultaneously we can have up to maybe 30 SLR receiving redo.

  For the SLR=primary redo+1, I guess they want to make sure that enough standby redos exsit to receive the primary's redo, so that even all primary redo logs are active, we still have one more SLR avaible. Image that "lgwr+sync" mode,  you can have one-one relation between primary log and standby log, which can provide the highest availability and performance. (ignore this part if you don't believe it ^_^)

Yong Huang

I was able to see the extra SRL group shown as ACTIVE in v$standby_log. The details are at

http://www.freelists.org/post/or ... by-redo-log-group,8

Yong Huang

amaome

Hi Yong, nice to see u again.

now i'm using 10g managed standby recovery.
On primary, there're 6 redo logs and 7 standby logfiles. However, on standby I increased standby logfile to 12.  
on standby,  v$standy_log

SQL> select group#, thread#, sequence#, used, archived,status
  2  from   v$standby_log;

    GROUP#    THREAD#  SEQUENCE#       USED ARC STATUS
---------- ---------- ---------- ---------- --- ----------
         7          1      99289 1064481792 NO  ACTIVE
         8          1      99302 1064627200 NO  ACTIVE
         9          1      99279 1060516864 NO  ACTIVE
        10          1      99297 1065680896 NO  ACTIVE
        11          1      99294 1059560960 NO  ACTIVE
        12          1          0        512 NO  UNASSIGNED
        13          1      99281 1059644928 NO  ACTIVE
        14          1      99285 1061145600 NO  ACTIVE
        15          1      99274 1065680896 NO  ACTIVE
        16          1      99286 1061635584 NO  ACTIVE
        17          1      99291 1062708736 NO  ACTIVE

    GROUP#    THREAD#  SEQUENCE#       USED ARC STATUS
---------- ---------- ---------- ---------- --- ----------
        18          1      99303 1062132224 NO  ACTIVE

one question about the alert*.log from standby as below:

RFS[5]: Successfully opened standby log 12: '/data2/dw1/redostby12.log'
Fri Apr 24 12:39:16 2009
Media Recovery Log /archive/dw1/1_97454_656126950.dbf
Fri Apr 24 12:40:16 2009
RFS[9]: No standby redo logfiles of size 2078026 blocks available
Fri Apr 24 12:41:21 2009
Media Recovery Log /archive/dw1/1_97455_656126950.dbf
Fri Apr 24 12:41:31 2009
RFS[6]: No standby redo logfiles of size 2077964 blocks available


do u know what's meaning of this clause ' No standby redo logfiles of size 2078026 blocks'?
Is it just friendly reminder, or some potential disadvantage?

backupground:　oracle10g204+AIX 5.3
redo log: 1GB
standby redo log: 1GB